**ETF Trading Strategies **

### By Bob Krishfield

### West Coast Cycles Club

### Jan 29, 2013

**Agenda **

### Strategies for ETF Trading

### –

### Concepts & Modeling

### –

### Examples

### –

### Testing

### –

### Advanced Models

### Part 2

### •

### Tools For ETF Trading

### –

### To Validate & Implement

### •

### Modeling and Validating in AmiBroker

**Trading Strategies **

### •

### Non-Systematic Trading a Looser for Average Investor

### •

### Lots of Tools Available, But Key is How to Apply them

### •

### Need to Choose and Follow a Strategy.

### •

### Apply Principles of Portfolio Investing

1. Design According to the Investor's Risk Tolerance

2. Allocate Assets to Maximize the Expected Return and Minimize the Risk

3. Manage Portfolio Utilizing Rules Validated Through Testing

### My Objective is To Identify Strategies and Then Cover Some Tools

### to Implement Them

**Portfolio Strategy Concepts **

### •

### Buy & Hold – A Passive Strategy

### –

### Portfolios with Diversified Assets Can Do Well

### •

### Strategic Allocation

### –

### Allocation Across Non-Correlated Assets/Markets*

### –

### Rebalance Monthly or Annually to Targets Improves Reward

### to Risk Ratio

### •

### Tactical Asset Allocation

### –

### Use Trading to Allocate Assets with Highest Potential

### Returns and Avoid Those with Potential Losses

### –

### Select from List of Strategically Selected Assets

1/29/13

**Portfolio Modeling **

### •

### Major Problem in Applying Modern Portfolio Theory* is the Use

### of Static, Long Term Averages for Allocations

– Allocations Not Updated as Market Changes

### •

### Solution: Evaluate and Apply these Parameters Monthly

### •

### Optimal Portfolios**

– Based on 4 Parameters:

• Momentum (Relative Strength)

• Valuation

• Volatility

• Correlation

– Require Strategic Diversification Across Asset Classes*.*

• *Markowitz, “Modern Portfolio Theory”, 1967

**Momentum – Relative Strength **

1/29/13

37 Randomly Selected ETFs in 2010

70% of Variablility in Return Explained by Rel Strength 83% Correlation

**Returns and Risk **

### •

### Momentum Only Include Assets Rising in Value

### •

### Value* Select Assets That Are on Low Range of Valuation

### (i.e Range of OverBought – OverSold)

### •

### Market Risk ( price volatility )

– Diversify: Pick Non-Correlated Markets

– Diversify - Use ETFs (basket of stocks)

– Reduce Uncertainty - Avoid New ETFs

### Measures

– Compounded Annual Return (CAR)

– Volatility

• Measured as Annualized Standard Deviation

– Sharpe Ratio (simplified)

• Measure of Risk Adjusted Return. Good if Above 1.0

• Portfolio CAR / Standard Deviation

– Drops Out Risk Free Return Which Currently is About 0.0

### Classes of Models Presented

1. Fixed Allocation Across

Strategic Asset Classes 2. Fixed Allocation with Timing

3. Relative Strength

(Momentum) Allocation, Equal Weighted

ü With Filters for Crash

Protection

4. Relative Strength Allocation

with Dynamic Weighting

ü Volatility Weighted

ü Minimum Variance Weighting

**Portfolio Strategies for ETF Trading **

1/29/13

### Example Portfolios by Class

1. Permanent Portfolio

All-Weather ETF Portfolio 2. Faber’s Ivy Portfolio

3. Relative Str Ranking

Kennedy’s Coolcat ETF Model R W Colby”s Top 10 ETF Model Masonson’s Buy-Don’t-Hold

Model Performance will be compared using results from Testing, since

**1 - Fixed Portfolio Example #1 **

**Model Definition **

### •

### Portfolio Setup from Mix of ETFs That Would Be Safe and

### Profitable in Any Kind of Economic Cycle

### •

### Harry Brown’s, Permanent Portfolio uses 25% for Each of 4

### ETFs – One for Each Part of Economic Cycle

### •

### Buy & Hold, Rebalance Annually

Growth Stocks for Prosperity VTI Precious Metals/Gold for Inflation GLD Govt Bonds for Deflation TLT Cash(ST Bonds) for Recession SHY

**Results **

– **Returned 9.5% over last 40 years**.

– **Mutual Fund (PRPFX) 3yr CAR = 10.6 **

AAII, “The Permanent Portfolio / Using Allocation to Build and Protect Wealth”, 2011

**1 - Fixed Portfolio Example #2 **

### •

### Portfolio Setup to Perform Well

### Over All Market Environments

Ray Dalio, Bridgwater Assoc. Setup Strategy. His landmark concept was to create a portfolio that would have

roughly **equal risk **in four different
economic regimes:

1) rising growth

2) falling growth 3) rising inflation 4) falling inflation.

Same As Permanent Portfolio Concept But Different Assets Used.

1/29/13

**1 - Fixed Portfolio Examples **

### Portfolio to Have Equal Risk Over Market Environments Setup

### Using Weighted Fixed Allocations Based on Risk

Results for 2005-2012: CAR=8.1%, Sharpe=.83, MDD=-19.7, V=6.1%

VTI EMB DBC,GLD HYG TLT,IEF TIP

**1 - Fixed Portfolio Comparison **

### •

### Fixed 4 Portfolio Setup from Strategic Mix of ETFs That Would

### Be Safe and Profitable in Any Kind of Economic Cycle, Rebal

### Annually.

### •

### ETFs Same as Perm Portfolio: VTI, TLT, GLD, SHY

### •

### Model Testing Using ETFReplay 2007-13

1/29/13

**Test Model **

**Model ** **CAR ** **Sharpe ** **MDD ** **Volatility **

Permanent 4 ETFs 40 yrs 9.5

Permanent Mut Fund 3 yrs 10.3 All-Weather 6 ETFs 05-12 8.1 .83 -19.7 6.1 Fixed 4 Test Model 07-13 13.4 .78 -3.5 6.8

### •

### Setup List of a Strategic Mix of up to 10 Asset Class ETFs

### •

### Apply Market Timing to Reduce Drawdown

– Update Monthly – Buy/Sell If Above/Below 200 day

### •

### Example – Faber’s Ivy Portfolio

IVY Portfolio WebsiteTwo Portfolio Implementations In Book

**2 - Fixed Portfolio with Timing **

**Model Definition **

**2 - Fixed Portfolio with Timing **

### Faber’s Performance Using 10 Mo Timing

– Ivy5 Portfolio 1973-2008 CAR= 11.27%, V= 6.87,MDD=-9.5%,S=.77

Benchmark B&H No Timing CAR= 9.77%, V=9.73,MDD=-36%,S=.39

### Model Runs 2007-13 Using ETFReplay.com

1/29/13

Results Not Great, But Model Simple – Inspired Many to Improve Model

**Model ** **CAR Sharpe ** **MDD ** **Volatility **

1-Fixed 4 ETFs (Perm4) 13.4 .78 -3.5 6.8 1-Fixed 5 ETFs (Ivy5) 11.0 .30 -7.1 12.1 1-Fixed 10 ETFs (Ivy10) 10.7 .23 -8.3 13.6

**2-Fixed 5 (Ivy5) w Timing4 ** **9.1 ** **.59 ** **-13.2 ** **10.2 **

**3 - Relative Strength Asset Allocation **

### •

### Setup List of a Strategic Mix of up to 10 Asset Class ETFs

### •

### For Testing & Comparison – Use IVY10 List

### •

### Relative Strength Used to Allocate Assets to Portfolio on a

### Monthly Basis, .i.e Pick Best Performers

### •

### Use Equal Weighting

### •

### Buy Top x% from Ranked List Using N month Momentum (Rel

### Strength)

### •

### Optional –

– Crash Proof Filter Sell/Don’t Buy If Rel Strength is Zero or Neg

– Alternative – Don’t Buy if Rel Stength is Less Than Prior Month

### •

### Rel Strength Model Testing

### 2007 – 2013

– Rel Str Lookback = 4 - 6 months,

– Select 2 or 3 from 10, Updated Monthly

1/29/13

**3 - Relative Strength Asset Allocation **

**Model Testing **

**Model ** **CAR ** **Sharpe ** **MDD ** **Volatility **

1-Fixed 4 (Perm4) 13.4 .78 -3.5 6.8 1-Fixed 5 (Ivy5) 11.0 .30 -7.1 12.1 1-Fixed 10 (Ivy10) 10.7 .23 -8.3 13.6 2-Fixed 5 (Ivy5) w Timing 9.1 .59 -13.2 10.2 2-Fixed 10 (Ivy10) w Timing 8.4 .53 -12.9 10.3

**3-RelStr4 2-10 wCP ** 21.8 .96 -17.3 19.1

**Results from My Testing* **

### •

### Relative Strength Ranking Better than Fixed B&H

– Lookback **Period of 4 months **generally worked best

– Use of 2 Rel Str Periods in Ranking

• Mixed Results. Better if Only 1-2 Picks, Worse for 3-5 Picks – Monthly Timing & Portfolio Updates

• Semi-Monthly Failed to Improve Performance in Most Cases

### •

**Crash Protection**

### Filter – definite improvement

– Returns Not always Higher, but Volatility Lower with Higher Sharpe

### •

### Volatility Used for Ranking

– ETFReplay Ranking Model – Optional Weighting of Volatilty

• Mixed Results. Better When Picking 4-5 from List, Worse Otherwise

• Volatility Tested with Equal Weight of Relative Strength.

### •

### Volatility Weighting vs Equal Weighting

– Tbd. Current Software Tools Not Capable.

– All Tests Used Equal Weighting

1/29/13

**CAR Ploted vs %Volatility for RS4 case **

CP (Crash Protection) Improved Performance by Reducing Volatility

**Use of Crash Protection Filtering **

5.0 10.0 15.0 20.0 25.0 30.0 35.0 10.0 12.0 14.0 16.0 18.0 20.0 22.0 24.0 26.0 RS-4 CP RS-4 Linear (RS-4 CP)

**Relative Strength Lookback Period **

0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
10.0 12.0 14.0 16.0 18.0 20.0 22.0 24.0 26.0 28.0
RS3*
RS3
rs4
RS5
RS6
Linear (RS3)
Linear (rs4)
Linear (RS5)
Various Rel Strength Periods Compared RS = 4 Months Superior

### •

### Setup List of a Strategic Mix of 10 Asset Class Funds

– US, Europe, Japan, Emerg Mkt Stocks, US & Intl REITS, Long & ST US

Treas, Commodities, Gold

### •

### Six Models Analyzed:

• Equal Weight

• Volatility Weighted

• Top 5 Momentum Ranked, Equal Weight

• Top 5 Momentum Ranked, Volatility Weighted

• Top 5 Having Minimum Volatility & Correlation

• Optimized & Integrated Use of Momentum, Volatility, Correlation

### •

### Tests Run over Period 2002-12

1/29/13

**4 - **

**Relative Strength w/ Dynamic Weighting **

**Model Definitions **

*Based on Paper by A. Butler,”Adaptive Asset Allocation – A Primer”, 2012

This Study* Provides Insight Into More Complex Models and What Might Be Gained From Their Use.

**4 - **

**Relative Strength w/ Dynamic Weighting **

**Model ** **CAR Sharpe MDD **

1.Equal Weight B&H, Rebal Mo 8.36 .66 -44 2.Equal Volatility Weight, Rebal Mo 8.88 1.23 -19 3.Top 5 Rel Str Rank, Equal Wt 14.3 1.23 -26 Models Requiring Specialized Software

4.Top 5 Rel Str Rank, Volatility Wt 13.9 1.53 -15.9 5.Top 5 Rel Str Rank, Minimum Variance 15.4 1.71 -15.8 6.Optimized (R,V,C) Weekly Rebalanced 16.9 2.51 -9.6

*Based on Paper by A. Butler,”Adaptive Asset Allocation – A Primer”, 2012

From 1995 to 2011 (16 Years)

Models 1-3 Similar to those Previously Presented.

Models 4-6 Show Potential for Improved Portfolio Performance But More Complexity, Requiring Specialized Software

**What’s Required for More Complex Models **

### •

### Models 4 thru 6 Achieve Superior Results But

### •

### Custom Software Required

### •

### Ranking Algorithm Must Use Data from all

### Rows to Compute Score for Each Row

### •

### Optimization Model required for Last Model

### •

### One Approach is the Use of Monte Carlo

### Methods

### •

### See Quantext by Gauss

### •

*evidence suggests that a simple asset class ETF *

*momentum strategy beats an equal-weighted *

*benchmark over the past nine years under *

*conservative assumptions, adding value to simple *

*diversification... *

### •

*evidence indicates that strategies setting *

*portfolio-level volatility targets for a set of global stock indexes *

*easily outperform equal weighting based on gross *

*Sharpe ratio and maximum drawdown. *

* CXOAdvisory.com*

**References **

1. M. Faber, “A Quantitative Approach to Tactical Asset Allocation”, 2009

2. H. Browne, “Fail-Safe Investing: Lifelong Financial Security in 30 Minutes”, 1999

3. W. ThorpeAAII, “Using Relative Strength Analysis to Invest in ETFs”, 2011

4. R. Colby, “Introduction to the Screening Method for Analysis of Relative Strength”,2009

5. Seeking Alpha, “All Weather ETF Portfolio”

6. Bridgwater Assoc., “The All Weather Story”

7. K. Kennedy Coolcat.com, “The Coolcat ABCs of ETFs”, 2005

8. L. Masonson, “Buy Don’t Hold: Investing with RTFs Using Relative Strength to Increase

Returns with Less Risk”, 2010

9. L. Herr, “Modeling the Faber-Richardson Ivy Portfolio, 2012

10. A Butler, “Adaptive Asset Allocation – A Primer”, 2012

11. W. Keller,”Generalized Momentum and Flexible Asset Allocation, 2012

12. Gauss, “Reducing Risk Using Modified Ivy Portfolio”,2011

13. CXOAdvisory, “Asset Allocation Combining Momentum, Volatility, Correlation and Crash

Protection”, 2012

14. CXOAdvisory, “Diversifying Across Strategic Allocation Strategies”, 2012

15. CXOAdvisory, “Combining SMA Crash Protection and Momentum in Asset Allocation, 2012